#!/bin/bash
#grab container image name
MODEL_NAME="Qwen3-Coder-Next"
CONT_NAME="ik-$MODEL_NAME"

MODEL="Qwen3-Coder-Next-IQ4_KSS.gguf"
MODEL_FOLDER="/sam4t/ENC/LLM/Qwen3-Coder-Next"

source ./ik_env.sh
echo $IK_IMAGE
docker container create --name $CONT_NAME --network llms -p 10987:10987 -v $MODEL_FOLDER:/model --user $(id -u):$(id -g) --gpus=all --restart on-failure:3 \
	--health-cmd "curl -f http://$CONT_NAME:10987/v1/models || exit 1" \
	--health-interval 5s \
	--health-timeout 2s \
	--health-start-period 15s \
	$IK_IMAGE \
	/ik_llama.cpp/build/bin/llama-server \
	--model /model/$MODEL \
	-ctk q8_0 -ctv q8_0 \
	--parallel 1 --threads 32 \
	--host $CONT_NAME \
	--port 10987 -ngl 99 \
	--alias $MODEL_NAME \
	--no-mmap --jinja \
	-b 4096 -ub 4096 \
	-ns 1 -np 1 -cram 32000 \
	-cram-n-min 128 -crs .9 \
	--ctx-checkpoints 128